Skip to content

ooda-and-execution.md

Latest commit

 

History

History
70 lines (53 loc) · 1.75 KB

ooda-and-execution.md

File metadata and controls

70 lines (53 loc) · 1.75 KB
Raw

OODA and Execution Flow

This document maps Loop runtime behavior to Observe/Orient/Decide/Act.

Observe

Inputs gathered:

  • User intent
  • Session context
  • File/tool outputs
  • Prior memory and trace data (when relevant)

Key modules:

  • context
  • memory (read paths)
  • adapters::claude_code::adapter (request-context ingestion + memory lookup)

Orient

Analysis performed:

  • Complexity signal extraction
  • Domain/workflow inference
  • Constraint and budget awareness

Key modules:

  • complexity
  • adapters::claude_code::hooks (prompt analysis and signal propagation)
  • spec_agent::parser (for NL formalization flows)

Decide

Decision points:

  • Activate or bypass recursive orchestration
  • Select model/routing path
  • Select formalization depth and verification posture
  • Select fallback behavior under failure/limits

Key modules:

  • orchestrator
  • llm::router
  • adapters::claude_code::adapter (activate/skip + mode selection bridge)
  • signature and fallback extraction paths

Act

Actions executed:

  • Run REPL/module operations
  • Perform LLM calls
  • Emit trajectory events
  • Persist memory/traces
  • Run verification and governance checks

Key modules:

  • repl
  • adapters::claude_code::adapter (execution + metadata shaping)
  • module
  • reasoning
  • dp governance wrapper (./scripts/dp)

Observability Expectations

A healthy execution path produces:

  1. Traceable decisions.
  2. Deterministic command evidence for gates.
  3. Recoverable diagnostics when failures occur.
  4. Scenario-level adapter efficacy evidence (VG-CLAUDE-ADAPTER-E2E-001), not just plumbing tests.

If one of these is missing, the system may still run, but operating it safely becomes mostly archaeology.

For a full request lifecycle trace, read runtime-walkthrough.md.